CN110781811A

CN110781811A - Abnormal work order identification method and device, readable storage medium and computer equipment

Info

Publication number: CN110781811A
Application number: CN201911016245.2A
Authority: CN
Inventors: 张婧琦; 邹耿鹏
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2019-10-24
Filing date: 2019-10-24
Publication date: 2020-02-11

Abstract

The application relates to an abnormal work order identification method, an abnormal work order identification device, a readable storage medium and computer equipment, wherein the method comprises the following steps: acquiring a work order image in a current reported work order, and acquiring work order image characteristics corresponding to the work order image according to the work order image; performing resource transfer operation detection on the work order image according to the work order image characteristics to obtain a first detection result; performing black friend operation detection on the work order image according to the work order image characteristics to obtain a second detection result; and when determining that the resource transfer operation exists in the work order image of the current report work order according to the first detection result and determining that the black-out friend operation exists in the work order image of the current report work order according to the second detection result, determining that the current report work order is an abnormal work order. According to the method, the image provided in the reported work order is preliminarily judged through image processing and a computer, the reported work order can be processed in an auxiliary mode, and the processing efficiency of the reported work order is improved.

Description

Abnormal work order identification method and device, readable storage medium and computer equipment

Technical Field

The present application relates to the field of computer technologies, and in particular, to a method and an apparatus for identifying an abnormal work order, a computer-readable storage medium, and a computer device.

Background

With the development of technology, people can not only achieve the purpose of entertainment through the electronic device, but also establish contact with friends through the electronic device, and can also realize resource transfer and the like on the electronic device through a network, however, the functions provide convenience for people, and meanwhile, various phishing events occur.

Many software will report the center, when the user finds out the behavior that does not accord with the regulation in the course of using software, can adopt the screenshot to report the material and initiate to report the center as reporting usually. However, the report center usually needs to process a large number of report work orders, and the processing flow is to manually and sequentially examine the report materials in the report work orders according to the time sequence of receiving the report work orders, select report work orders which may have higher fraud suspicions from the report work orders, and give priority to customer service personnel for processing; thus, the processing efficiency for the report work order is low.

Disclosure of Invention

In view of the foregoing, it is desirable to provide an abnormal work order identification method, apparatus, readable storage medium and computer device.

An abnormal work order identification method, the method comprising:

acquiring a work order image in a current reported work order, and acquiring work order image characteristics corresponding to the work order image according to the work order image;

performing resource transfer operation detection on the work order image according to the work order image characteristics to obtain a first detection result;

performing black-drawing friend operation detection on the work order image according to the work order image characteristics to obtain a second detection result;

and when determining that the resource transfer operation exists in the work order image of the current report work order according to the first detection result and determining that the blacking friend operation exists in the work order image of the current report work order according to the second detection result, determining that the current report work order is an abnormal work order.

An abnormal work order identification apparatus, the apparatus comprising:

the image characteristic acquisition module is used for acquiring a work order image in a current reported work order and acquiring work order image characteristics corresponding to the work order image according to the work order image;

the first detection module is used for carrying out resource transfer operation detection on the work order image according to the work order image characteristics to obtain a first detection result;

the second detection module is used for carrying out blackening friend operation detection on the work order image according to the work order image characteristics to obtain a second detection result;

and the abnormal work order identification module is used for determining that the current report work order is an abnormal work order when determining that the resource transfer operation exists in the work order image of the current report work order according to the first detection result and determining that the operation of a blacking friend exists in the work order image of the current report work order according to the second detection result.

A computer-readable storage medium, storing a computer program which, when executed by a processor, causes the processor to perform the steps of the method as described above.

A computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of the method as described above.

The abnormal work order identification method, the abnormal work order identification device, the readable storage medium and the computer equipment are used for acquiring the work order image in the reported work order, respectively carrying out resource transfer operation and blacking friend operation detection on the work order image through image characteristics corresponding to the work order image to obtain a first detection result and a second detection result, and preliminarily judging whether the reported work order has high suspicion or not according to whether the resource transfer operation and the blacking operation exist in the detection results of the work order images at the same time so as to select the abnormal work order; the abnormal work order can be preferentially distributed to customer service staff for processing, the image provided in the report work order is preliminarily judged through processing the image and a computer, the report work order can be processed in an auxiliary mode, and the processing efficiency of the report work order is improved.

Drawings

FIG. 1 is a diagram of an exemplary application of the abnormal work order identification method;

FIG. 2 is a schematic flow chart diagram illustrating an abnormal work order identification methodology in one embodiment;

FIG. 3 is a schematic diagram of a model structure of VGG16 in one embodiment;

FIG. 4 is a diagram showing a structure of a model of fast R-CNN in one embodiment;

FIG. 5 is a flowchart illustrating an abnormal work order identification method according to an exemplary embodiment;

FIG. 6 (left) is a schematic diagram of a virtual red envelope interface in a resource transfer operation in an embodiment;

FIG. 6 (right) is a diagram illustrating a transfer details interface in a resource transfer operation in an exemplary embodiment;

FIG. 7 is a diagram of a chat interface in one embodiment;

FIG. 8 is a schematic illustration of a current report order output in an exemplary embodiment;

FIG. 9 is a block diagram showing the structure of an abnormal work order recognition apparatus according to an embodiment;

FIG. 10 is a block diagram showing a configuration of a computer device according to an embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

FIG. 1 is a diagram of an exemplary application of the abnormal work order identification method. Referring to fig. 1, the abnormal work order identification method is applied to an abnormal work order identification system. The abnormal work order recognition system includes a terminal 110 and a server 120. The server 120 obtains a work order image in the current report work order, obtains corresponding work order image characteristics, performs resource transfer operation and black-out friend operation detection on the work order image according to the work order image characteristics, obtains a first detection result and a second detection result, and determines that the current report work order is an abnormal work order when the resource transfer operation is determined to exist in the current work order according to the first detection result and the black-out friend operation is determined to exist in the current work order according to the second detection result. In some embodiments, the server 120 obtains the current report order from the terminal 110. The terminal 110 and the server 120 are connected through a network. The terminal 110 may specifically be a desktop terminal or a mobile terminal, and the mobile terminal may specifically be at least one of a mobile phone, a tablet computer, a notebook computer, and the like. The server 120 may be implemented as a stand-alone server or a server cluster composed of a plurality of servers.

As shown in FIG. 2, in one embodiment, an abnormal work order identification method is provided. The embodiment is mainly illustrated by applying the method to the server 120 in fig. 1. Referring to fig. 2, the abnormal work order identification method specifically includes steps S210 to S240.

And step S210, acquiring a work order image in the current report work order, and acquiring work order image characteristics corresponding to the work order image according to the work order image.

In this embodiment, the image in the report work order is recorded as a work order image. When the user reports, the user reports the work order image obtained in this embodiment by using a virtual red envelope interface, a transfer interface, a chat screenshot, or a screenshot of another interface, as a report material. The work order image characteristics corresponding to the work order image are obtained according to the work order image, and the work order image characteristics can be obtained in any mode.

In one embodiment, after acquiring the work order image in the currently reported work order, the method further comprises preprocessing the work order image, and acquiring the work order image feature from the work order image acquired after preprocessing.

The image preprocessing method and the image preprocessing device can filter some images which do not conform to a preset format, process the work order images into the same type of images and facilitate subsequent detection. For example, in one embodiment, after acquiring the work order image, the method further comprises: identifying the work order image to obtain an image format of the work order image; in this embodiment, the method for identifying an abnormal work order further includes the steps of: and converting the image format of the work order image into a preset image format.

The image format, that is, the format in which the image file is stored on the memory card, is usually JPEG, TIFF, RAW, or the like; in this embodiment, the image format of the work order image is identified after the work order image is obtained, and the work order image in the non-preset image format is converted into the work order image in the preset format. In the method, the work order images are converted into the same image format, so that detection errors caused by different image formats can be avoided; in one embodiment, the predetermined image format is a JPEG format.

In another embodiment, the work order image is identified, and the image size of the work order image is obtained; in this embodiment, the method for identifying an abnormal work order further includes the steps of: and filtering the work order image with the image size not in accordance with the preset proportion, and adjusting the image size of the work order image to be the preset size.

Wherein, the image size corresponds to the size of the work order image, and further, the method further comprises the following steps after the size of the work order image is obtained: in the embodiment, the image proportion of each work order image is detected, the work order images which do not accord with the preset proportion are filtered, and the sizes of the rest work order images are non-preset sizes and are adjusted to be the preset sizes; further, in one embodiment, the size of the image is adjusted by cubic interpolation or the like. The images which do not accord with the preset proportion are likely to be cut or spliced images which are likely to be tampered, so that in the method, the work order images which do not accord with the preset proportion are filtered, and the accuracy of the abnormal work orders can be identified.

In another embodiment, the work order image is identified, and an image channel of the work order image is obtained; in this embodiment, the method for identifying an abnormal work order further includes the steps of: the filtered image channel includes a work order image of an alpha channel.

In the embodiment, in the process of preprocessing the work order image, the work order image containing the Alpha Channel is filtered, and the work order image containing the Alpha Channel is likely to be a cut or spliced image which is possibly tampered, so that in the method, the work order image which does not accord with the preset proportion is filtered, and the accuracy of the abnormal work order can be identified.

In another embodiment, the work order image is identified, and image pixel points of the work order image are obtained; in this embodiment, obtaining the work order image feature corresponding to the work order image from the work order image includes: and obtaining the work order image characteristics corresponding to the work order image according to the image pixel points of the work order image. After the image pixel points of the work order image are obtained, three-dimensional data of RGB (Red, Green and Blue channel colors) of each image pixel point can be further obtained, namely the corresponding work order image characteristics.

Wherein, the pixel is composed of small squares of the image, the small squares have a definite position and assigned color value, and the color and the position of the small squares determine the appearance of the image; in this embodiment, after the work order image is obtained, the corresponding image pixel points are also obtained, and the work order image features are obtained through the image pixel points.

Further, in one embodiment, the work order image is processed by a Python image processing library (Python imaging library) to obtain image information of the work order image, such as image format, image size, image pixel points, and image channels.

And step S220, performing resource transfer operation detection on the work order image according to the work order image characteristics to obtain a first detection result.

Whether the resource transfer operation exists in the work order image is determined by analyzing the obtained work order image features, and in this embodiment, the result obtained by detecting the resource transfer operation is recorded as a first detection result. In one embodiment, the resource transfer operation is represented as a payment operation, such as: and when the virtual red packet is sent, the other party receives the virtual red packet or transfers on the internet and receives the transfer from the other party and the like are detected in the work order image, a first detection result of the resource transfer operation is obtained. It can be understood that if the operation of sending the virtual red envelope and the other party having received the virtual red envelope or the on-line transfer and the other party having received the transfer is not detected in the work order image according to the work order image features, a first detection result that there is no resource transfer will be obtained.

In one embodiment, the work order image features are analyzed by training the determined neural network model to achieve the purpose of detecting whether the resource transfer operation exists in the work order image.

For example, in a specific embodiment, when the classification model determined through training detects that the category of the work order image is resource transfer, it is determined that a resource transfer operation exists in the work order image, for example, when the classification model detects that the work order image corresponds to a virtual red envelope or a transfer screenshot, it is determined that the category of the work order image is resource transfer; for another example, in another specific embodiment, when it is detected that the work order image includes target information related to resource transfer through the trained target detection model, it is determined that a resource transfer operation exists in the work order image, for example, when it is detected that the work order image corresponds to a chat screenshot including a virtual red packet received by the other party or an account transfer through the target detection model, it is determined that the work order image includes the target information related to resource transfer. The classification model is obtained by performing model training on the basis of a historical work order image carrying classification labels; the target detection model is obtained by performing model training on a historical work order image based on a label carrying target information. The resource transfer detection of the work order image is completed by training the determined neural network model, so that the processing efficiency can be improved, and the accuracy of the detection result can be improved. It is understood that in other embodiments, the target information detection of the resource transfer operation may be implemented in other manners.

Step S230, performing black-out friend operation detection on the work order image according to the work order image characteristics to obtain a second detection result.

Whether the blacking friend operation exists in the work order image is determined by analyzing the obtained work order image features, and in the embodiment, the result obtained by detecting the resource transfer operation is recorded as a second detection result. In one embodiment, the pull black buddy operation appears to delete a buddy, pull a buddy into a blacklist, and so on. When the existence of a friend deleting operation (such as a prompt that 'you are not a friend of the other party yet') and the operation of pulling the friend into a blacklist (such as a prompt that 'a message is sent but rejected by the other party') are detected in the work order image, a second detection result of the existence of the operation of pulling the friend into the blacklist is obtained. It can be understood that if the operation of deleting a friend or pulling a friend into a blacklist is not detected in the work order image according to the work order image features, a second detection result without the existence of the black-pulled friend is obtained.

In one embodiment, the work order image features are analyzed through a neural network model determined through training, so that the purpose of detecting whether black-out friend operation exists in the work order image is achieved. In a specific embodiment, when it is detected by the trained and determined target detection model that the work order image includes target information related to the operation of the blacking-out friend, it is determined that the operation of the blacking-out friend exists in the work order image, and when it is detected by the target detection model that the work order image corresponds to the target information including "you are not a friend of the other party" and "the message has been sent but is rejected by the other party", it is determined that the work order image includes the target information related to the blacking-out friend. The target detection model is obtained by performing model training on the basis of a marked historical work order image carrying target information. The operation detection of the blackened friends of the work order image is completed by training the determined neural network model, so that the processing efficiency can be improved, and the accuracy of the detection result can be improved. In other embodiments, the target information detection of the black friend operation may also be implemented in other manners.

Step S240, when the resource transfer operation exists in the work order image of the current report work order according to the first detection result and the black-out friend operation exists in the work order image of the current report work order according to the second detection result, the current report work order is determined to be an abnormal work order.

When resource transfer operation and operation of a black friend exist in one report work order at the same time, the report work order is considered to have higher fraud suspicion, and the report work order can be judged as an abnormal work order at the moment.

In one embodiment, one reported work order only comprises one work order image, the work order image is subjected to resource transfer operation and black-drawing friend operation detection, and when the resource transfer operation and the black-drawing friend operation are detected in the work order image, the reported work order is determined to be an abnormal work order; in another embodiment, a plurality of work order images may exist in one reported work order, resource transfer and black friend operation needs to be performed on each work order image respectively, and then whether the reported work order is an abnormal work order is determined according to the first detection result and the second detection result of each work order image.

In an embodiment where the currently reported work order includes a plurality of work order images, the method for identifying an abnormal work order includes the steps of: acquiring a first work order image in a current report work order, acquiring a first work order image characteristic corresponding to the first work order image, performing resource transfer operation detection on the first work order image according to the first work order image characteristic to acquire a first detection result, performing black friend operation detection on the first work order image to acquire a second detection result, judging whether the current report work order is an abnormal work order according to the first detection result and the second detection result, if not, returning to the step of acquiring the work order image of the current report work order, acquiring the second work order image in the current report work order, and sequentially executing detection steps until a first detection result and a second detection result of the second work order image are acquired, at the moment, judging whether the current report work order is the abnormal work order according to the first detection result, the second detection result of the first work order image and the first detection result and the second detection result of the second work order image, until all the work order images in the current reported work order are detected.

According to the method, the resource transfer detection and the black friend detection are sequentially carried out on the work order image in the current report work order, and if the current report work order is determined to be an abnormal work order according to the detection result of the detected work order image, the detection on other work order images is not needed, so that the efficiency of identifying the abnormal work order can be improved.

It can be understood that, in the embodiment of the present application, when the resource transfer operation can only be determined to exist in the current reported work order according to the first detection result, or when the blackout friend operation can only be determined to exist in the current reported work order according to the second detection result, it cannot be determined that the current reported work order is an abnormal work order.

Furthermore, after the current report work order is determined to be an abnormal work order according to the first detection result and the second detection result, the abnormal work order setting identification is set for the current report work order, the current report work order can be further picked out and output to a special channel and sent to exclusive customer service staff for processing, so that the report work order with high suspicion can be rapidly processed, and the efficiency of processing the report work order is improved.

The abnormal work order identification method comprises the steps of obtaining a work order image in a reported work order, respectively carrying out resource transfer operation and black-drawing friend operation detection on the work order image through image characteristics corresponding to the work order image to obtain a first detection result and a second detection result, and preliminarily judging whether the reported work order has high suspicion or not according to whether the resource transfer operation and the black-drawing operation exist in the detection results of all the work order images or not so as to select the abnormal work order; the abnormal work order can be preferentially distributed to customer service staff for processing, the image provided in the report work order is preliminarily judged through processing the image and a computer, the report work order can be processed in an auxiliary mode, and the processing efficiency of the report work order is improved.

In one embodiment, after obtaining the work order image feature corresponding to the work order image according to the work order image, the method further includes: and classifying the work order images according to the work order image characteristics to determine the category of the work order images. In this embodiment, performing resource transfer operation detection on the work order image according to the work order image feature, and obtaining the first detection result includes: when the type of the work order image is determined to be the resource transfer, a first detection result of the resource transfer operation is obtained.

In one embodiment, the categories into which the work order image may be classified according to its content include: chat, resource transfer (virtual red envelope or transfer) or other categories. In general, a user reports by intercepting an image containing evidence information as a report material (the work order image), wherein the intercepted image may be chat information content, a virtual red envelope or a transfer interface, or may also be other content, and corresponds to category chat, resource transfer or the like; in this embodiment, when the classification of the work order image is detected as a resource transfer, it is determined that there is a resource transfer operation in the currently reported work order.

Further, in another embodiment, the performing, according to the work order image feature, the resource transfer operation detection on the work order image, and obtaining the first detection result includes: when the category of the work order image is determined to be chat, performing target detection on the work order image according to the work order image characteristics; when the fact that the target information related to the resource transfer is contained in the work order image is detected, a first detection result of the resource transfer operation is obtained.

After the work order image is classified according to the work order image characteristics, if the obtained classification is chat, whether the resource transfer operation exists in the work order image can be further detected in a target detection mode, and when target information related to the resource transfer is detected, the resource transfer operation exists in the work order image is judged. When sending the virtual red packet or transferring the account to the other party, if the other party receives the virtual red packet or transfers the account, a prompt message, such as 'red packet received/transfer account' and the like, exists in the chat interface, and in this case, the user can also use the screenshot of the chat interface as the report material (namely, the work order image); in this embodiment, when the category of the work order image is determined to be chat, target detection is performed on the work order image according to the work order image characteristics, and when target information associated with resource transfer, such as ". dot.red packet/transfer", is detected, it is determined that a resource transfer operation exists in the work order image.

In one embodiment, after obtaining the work order image feature corresponding to the work order image according to the work order image, the method further includes: and classifying the work order images according to the work order image characteristics to determine the category of the work order images.

In this embodiment, performing blacking friend operation detection on the work order image according to the work order image feature, and obtaining the second detection result includes: when the category of the work order image is determined to be chat, performing target detection on the work order image according to the work order image characteristics; and when the fact that the work order image contains target information associated with the black-pulling friend is detected, obtaining a second detection result of the operation of the black-pulling friend.

After the friend is deleted or the friend is pulled into the blacklist by the other party, the message cannot be normally sent when the message is sent to the other party, and a prompt message related to the deletion or the pulling of the friend into the blacklist by the other party appears in the chat interface, such as 'not being the friend of the other party', 'the message is sent out, and the message is rejected by the other party', and the like, in this case, the user can intercept the image of the chat interface as a report material (namely, the work order image); in this embodiment, when the type of the work order image is determined to be chat, the target detection is performed on the work order image according to the work order image characteristics, and when target information associated with the blacking friend operation, such as "message sent and rejected by the other party", is detected, it is determined that the blacking friend operation exists in the work order image.

In the above embodiment, after the work order image features corresponding to the work order image are obtained, before the work order image is detected according to the work order image features, the work order image is classified, and the resource transfer operation detection and the black friend operation detection on the work order image are executed in a corresponding manner according to different classification results.

Further, in one embodiment, the scheme of the application relates to technologies such as an artificial neural network in machine learning of artificial intelligence, wherein the machine learning is a multi-field cross subject and relates to a plurality of subjects such as probability theory, statistics, approximation theory, convex analysis and algorithm complexity theory. The special research on how a computer simulates or realizes the learning behavior of human beings so as to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve the performance of the computer. In some embodiments of the application, the classification model and/or the target detection model obtained through machine learning training are used for detecting resource transfer operation and black friend operation in the work order image, and the identification accuracy can be improved. The following examples are intended to illustrate in particular:

in one embodiment, classifying the work order image according to the work order image features includes: inputting the characteristics of the work order images into a classification model determined through training, and classifying the work order images according to the classification model; the classification model is obtained by obtaining a sample work order image carrying classification labels in a historical report work order and performing model training on a preset neural network model according to the sample work order image carrying the classification labels.

The preset neural network is a classification network model, and in a specific embodiment, the preset neural network model is VGG16, as shown in fig. 3, which is a schematic diagram of a model structure of VGG16 in an embodiment. The training process for the VGG16 includes the steps of: acquiring work order images in historical reported work orders to construct training data; wherein, the historical work order image is classified manually in advance and labeled with category information, such as chatting, resource transfer (including virtual red envelope and transfer) or other categories; and then inputting the training data into a preset VGG16 network model for training, and stopping training when a termination condition is reached to obtain a final classification model. Wherein, the termination condition can be set according to the actual situation.

Further, in one embodiment, the output of the classification model is an image carrying an identifier of a "resource transfer," "chat," or "other" category, and the first detection result is determined based on the identifier carried in the output.

In one embodiment, there may be a plurality of result image classification results of the classification model, and each output result corresponds to a confidence level, and in this embodiment, the method further includes the steps of: and determining the classification result with the confidence coefficient higher than the threshold value as the final classification result of the work order image.

In another embodiment, the performing the resource transfer operation detection on the work order image according to the work order image feature, and obtaining the first detection result includes: and performing target detection on the work order image according to the work order image characteristics, and acquiring a first detection result of the resource transfer operation when determining that the work order image contains target information related to the resource transfer. Performing blacking friend operation detection on the work order image according to the work order image characteristics, and obtaining a second detection result comprises: and performing target detection on the work order image according to the work order image characteristics, and when the fact that the work order image contains target information associated with the black-pull friend is detected, obtaining a second detection result of the operation of the black-pull friend.

In this embodiment, whether a resource transfer operation or a blacking friend operation exists in the work order image is directly determined through target detection on the work order image. Taking the example of realizing operation detection on the work order image through a target detection model determined by training, acquiring historical work order images in a historical report work order to construct training data when the target detection model is trained; target areas are marked on each historical work order image in the training data, wherein the target areas comprise a target area related to resource transfer operation and a target area related to black friend operation; target areas of keywords such as 'virtual red envelope received by the opposite party' and 'transfer received' can be marked for the virtual red envelope or transfer screenshot in the historical work order image, and operations (resource transfer/blacking friends) corresponding to the target areas; for the historical work order image containing the prompt information related to the blacking friends and the resource transfer operation in the chat screenshot, the target area where the prompt information is located and the operation (resource transfer/blacking friends) corresponding to the prompt information are marked. And training a preset neural network model according to the training data to obtain a final target detection model.

In an embodiment, there may be a plurality of target region results output by the target detection model, and each output result corresponds to one confidence level, and in this embodiment, the method further includes the steps of: and determining the target area result with the confidence coefficient higher than the threshold value as the final result output by the target detection model, and determining the second detection result of the work order image according to the final result.

Further, in one embodiment, the performing the target detection on the work order image according to the work order image feature comprises: inputting the characteristics of the work order image into a target detection model determined through training, and carrying out target detection on the work order image according to the target detection model; the target detection model is obtained by obtaining a sample work order image carrying target information labels in a historical report work order and performing model training on a preset neural network model according to the sample work order image carrying classification labels.

The preset neural network is a classified network model, and in a specific embodiment, the preset neural network model is Faster R-CNN, as shown in fig. 4, which is a model structure diagram of Faster R-CNN in an embodiment. The training process for Faster R-CNN includes the steps of: acquiring work order images in historical reported work orders to construct training data; wherein, the training data is manually marked with a target information area in advance, such as a resource transfer operation prompt message in a chat screenshot, a prompt message pulled into a blacklist and the like; and then inputting the training data into a preset Faster R-CNN network model for training, and stopping training when a termination condition is reached to obtain a final target detection model. Wherein, the termination condition can be set according to the actual situation.

Further, in one embodiment, the result output by the target detection model is an image carrying the identifier of "resource transfer" or "blacking friends", and a second detection result is determined according to the identifier carried in the output result; specifically, the result output by the target detection model carries a resource transfer identifier, and a second detection result with resource transfer operation is obtained; the result output by the target detection model carries the identifier of the 'black friend', and a second detection result of the operation of the black friend is obtained; and the result output by the target detection model carries the resource transfer and blacking friend identification, and a second detection result of the resource transfer and blacking friend operation is obtained.

In one embodiment, the scheme provided by the embodiment of the application also relates to technologies such as image recognition, OCR and the like in computer vision of artificial intelligence; computer vision is a science for researching how to make a machine "see", and further, it means that a camera and a computer are used to replace human eyes to perform machine vision such as identification, tracking and measurement on a target, and further image processing is performed, so that the computer processing becomes an image more suitable for human eyes to observe or transmitted to an instrument to detect. As a scientific discipline, computer vision research-related theories and techniques attempt to build artificial intelligence systems that can capture information from images or multidimensional data. In some embodiments of the application, when the result output by the target detection model is that target information related to the operation of the black-out friend is detected, the work order image may be further identified through OCR. The following examples are intended to illustrate in particular:

in one embodiment, the target detection is performed on the work order image according to the work order image feature, and after it is detected that the work order image includes target information associated with the black-out friend, before a second detection result indicating that the black-out friend operation exists is obtained, the method further includes: and inputting the work order image into a character recognition model, judging that the work order image has the operation of the black-out friend when the work order image contains a preset keyword associated with the operation of the black-out friend according to the character recognition model, and obtaining a second detection result of the operation of the black-out friend.

In the process of chatting between the user and the opposite party, the system can also detect that the opposite party is suspicious (for example, the detection that the opposite party is reported by multiple persons or sensitive characters are detected in messages sent by the opposite party and the like), a prompt message can be sent to the user to prompt the user for risks, at the moment, a risk prompt message (for example, "please improve the alert" and the like) can appear on a chatting interface, the prompt message is similar to target messages related to the black-pulling friends, such as "the message sent but rejected by the opposite party" prompted by the chatting interface, and the like, so that in the embodiment, when the target messages related to the black-pulling friends are detected through a target, the work order image is further input into a character recognition model to carry out character recognition to determine whether the black-pulling friends exist in the work order image or not; specifically, after the work order image is input into the character recognition model, a character recognition result is obtained, and when a preset keyword related to a blacking friend is detected in the character recognition result, it is judged that the blacking friend operation exists in the work order image. The preset keywords related to the black-pulling friends can be set according to actual conditions.

The character recognition model is used for recognizing characters appearing in the image; in one embodiment, OCR (Optical Character Recognition) is used to recognize the work order image. OCR refers to a process in which an electronic device (e.g., a scanner or digital camera) examines a character printed on paper, determines its shape by detecting dark and light patterns, and then translates the shape into computer text using character recognition methods; the method is characterized in that characters in a paper document are converted into an image file with a black-white dot matrix in an optical mode aiming at print characters, and the characters in the image are converted into a text format through recognition software for further editing and processing by word processing software. In other embodiments, character recognition of the work order image may be accomplished in other ways.

According to the abnormal work order prompting method, for the detection of the operation of the blackened friends, which possibly has an uncertain detection result, after the target information related to the blackened friends is detected through the target, the work order image is input into the character recognition model to recognize the characters in the work order image, and whether the operation related to the blackened friends exists or not is further judged, so that the recognition accuracy of the abnormal work order can be improved.

In a specific embodiment, the detailed description is given by taking the example of implementing the detection of the work order image by training a determined classification model and a target detection model, wherein the classification model adopts a VGG16 model, and the target detection model adopts a Faster-R-CNN target detection model. Fig. 5 is a schematic flow chart of the abnormal work order identification method in this embodiment.

Acquiring a work order image in a current reported work order to obtain s1, s2, … and sn; preprocessing the work order image si, specifically comprising: reading an image, and acquiring information of the image, wherein the method comprises the following steps: image type, image size, image channel and image pixel point; converting the image in a non-jpg (png, jpeg, etc.) format into a jpg format; filtering out the work order images with alpha channels and the work order images with the size proportion not in accordance with the preset proportion; and adjusting the image to a preset size by using methods such as cubic interpolation and the like, and obtaining corresponding work order image characteristics v-feature.

Inputting the work order image feature v-feature into a classification model, and judging the category of the work order image; and when the type of the work order image is determined to be the virtual red envelope/account transfer, judging that the resource transfer operation exists in the work order image, and obtaining a first detection result of the resource transfer operation. Fig. 6 (left) is a schematic diagram of a virtual red envelope interface in the resource transfer operation in an embodiment, and fig. 6 (right) is a schematic diagram of a transfer detail interface in the resource transfer operation in an embodiment. In one embodiment, after obtaining the first detection result that the resource transfer operation exists, an identifier of 'resource transfer' is set for the current reported work order.

When the category of the work order image is determined to be chat, inputting the features of the work order image into a target detection model; when target information related to the resource transfer operation is detected in the work order image, such as 'red packet received' or 'transfer received', a first detection result that the resource transfer operation exists in the work order image is obtained. FIG. 7 is a diagram illustrating a chat interface including target information associated with a resource transfer operation.

When target information related to the operation of the black-pulling friend is detected in the work order image, such as 'pull-in blacklist' or 'delete friend' and the like, the work order image is input into an OCR character recognition model to obtain a character recognition result, and when the fact that the character recognition result of the work order image contains a preset keyword related to the operation of the black-pulling friend is detected, a second detection result of the work order image with the operation of the black-pulling friend is obtained. Fig. 7 is a schematic diagram of a chat interface including target information related to a blacking friend operation. In one embodiment, after a first detection result of the resource transfer operation is obtained, an identifier of a blacking friend is set for a current reported work order.

Further, in one embodiment, there may be a plurality of output results output by the classification model and the target detection model, each output result corresponds to a confidence level, and the output result with the confidence level higher than the threshold value is selected as the final result output by the model.

And when the category of the work order image is determined to be other categories, returning to the step of inputting the features of the work order image into the classification model, and inputting the image features of a second work order image of the currently reported work order into the classification model until a first detection result and a second detection result of each work order image are obtained.

Judging according to a first detection result and a second detection result of each work order image, if the first detection result is that resource transfer operation exists and the second detection result is that blacking friend operation exists, namely two operation behaviors exist in the current report work order at the same time, considering that the current report work order is abnormal, and determining the current report work order as an abnormal work order; in one embodiment, the "smart fraud" result is output in the current reported work order. Fig. 8 is a schematic diagram of a current report order output in an exemplary embodiment.

Wherein, training to obtain a classification model comprises the following steps: the training data is constructed by obtaining the historical work order images of the historical reported work orders, and the historical work order images are manually divided into resource transfer, chatting or other categories in advance and are labeled correspondingly. Training is carried out by using a VGG16 image classification algorithm based on training data carrying class marking information, and an image classification model is obtained.

The training to obtain the target detection model comprises the following steps: training data is constructed by obtaining historical work order images of historical reported work orders, the historical work order images are labeled manually in advance, target areas including resource transfer related prompt information and blacking friend related prompt information are used as the training data, and multiple target areas may exist in one image. Training the training data carrying the target information label by using a Faster R-CNN target detection algorithm to obtain a target detection model.

By the abnormal work order identification method, the system marking is provided, the time spent on manual evidence obtaining can be reduced, and the reporting work order with high fraud suspicion can be found without spending a large amount of time on manually checking all images; the work order image is detected through the neural network determined by training, so that the efficiency of processing the reported work order is greatly improved; after the abnormal work orders with high suspicion are selected, the abnormal work orders are preferentially sent to manual processing, so that the risk of delay penalty can be reduced; the system marking accuracy is high, manual secondary picture checking is not needed, and manual processing efficiency is improved.

Fig. 2 and 5 are respectively schematic flow diagrams of an abnormal work order identification method in an embodiment. It should be understood that although the steps in the flowcharts of fig. 2 and 5 are shown in sequence as indicated by the arrows, the steps are not necessarily performed in sequence as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 2 and 5 may include multiple sub-steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of performing the sub-steps or stages is not necessarily sequential, but may be performed alternately or alternately with other steps or at least some of the sub-steps or stages of other steps.

An abnormal work order recognition apparatus, as shown in fig. 9, the apparatus comprising: an image feature acquisition module 910, a first detection module 920, a second detection module 930, and an abnormal work order identification module 940.

The image feature obtaining module 910 is configured to obtain a work order image in the currently reported work order, and obtain a work order image feature corresponding to the work order image according to the work order image.

The first detecting module 920 is configured to perform resource transfer operation detection on the work order image according to the work order image feature, so as to obtain a first detection result.

The second detecting module 930 is configured to perform blacking friend operation detection on the work order image according to the work order image feature, so as to obtain a second detection result.

And the abnormal work order identification module 940 is configured to determine that the current reported work order is an abnormal work order when it is determined that the resource transfer operation exists in the work order image of the current reported work order according to the first detection result and it is determined that the blacking friend operation exists in the work order image of the current reported work order according to the second detection result.

The abnormal work order identification device acquires a work order image in the reported work order, respectively performs resource transfer operation and blacking friend operation detection on the work order image through image characteristics corresponding to the work order image to obtain a first detection result and a second detection result, and can preliminarily judge whether the reported work order has high suspicion or not according to whether the resource transfer operation and the blacking friend operation exist in the detection results of all the work order images simultaneously, so that the abnormal work order is selected; the abnormal work order can be preferentially distributed to customer service staff for processing, the image provided in the report work order is preliminarily judged through processing the image and a computer, the report work order can be processed in an auxiliary mode, and the processing efficiency of the report work order is improved.

In one embodiment, the above abnormal work order recognition apparatus further includes: the image classification model is used for classifying the work order images according to the work order image characteristics and determining the category of the work order images;

the first detection module is further used for obtaining a first detection result of the resource transfer operation when the type of the work order image is determined to be the resource transfer.

In one embodiment, the first detection module is further configured to perform target detection on the work order image according to the work order image feature when the category of the work order image is determined to be chat; when the fact that the target information related to the resource transfer is contained in the work order image is detected, a first detection result of the resource transfer operation is obtained.

In one embodiment, the second detection module is further configured to perform target detection on the work order image according to the work order image feature when the category of the work order image is determined to be chat; and when the fact that the work order image contains target information associated with the black-pulling friend is detected, obtaining a second detection result of the operation of the black-pulling friend.

In one embodiment, the first detection module is further configured to perform target detection on the work order image according to the work order image feature, and obtain a first detection result of the resource transfer operation when it is determined that the work order image includes target information associated with the resource transfer;

and the second detection module is further used for carrying out target detection on the work order image according to the work order image characteristics, and obtaining a second detection result of the operation of the black-pulling friend when the fact that the work order image contains target information associated with the black-pulling friend is detected.

In one embodiment, the above abnormal work order recognition apparatus further includes: and the character recognition module is used for inputting the work order image into the character recognition model, judging that the work order image has the operation of the black-pulling friend when the work order image contains the preset keywords associated with the operation of the black-pulling friend according to the character recognition model, and obtaining a second detection result of the operation of the black-pulling friend.

FIG. 10 is a diagram illustrating an internal structure of a computer device in one embodiment. The computer device may specifically be the server in fig. 1. As shown in fig. 10, the computer apparatus includes a processor, a memory, a network interface, an input device, and a display screen connected through a system bus. Wherein the memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program that, when executed by the processor, causes the processor to implement the abnormal work order identification method. The internal memory may also have a computer program stored therein, which when executed by the processor, causes the processor to perform the abnormal work order identification method. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.

Those skilled in the art will appreciate that the architecture shown in fig. 10 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.

In one embodiment, the abnormal work order recognition apparatus provided herein may be implemented in the form of a computer program that is executable on a computer device such as that shown in fig. 10. The memory of the computer device may store various program modules constituting the abnormal work order identification apparatus, such as the image feature acquisition module, the first detection module, the second detection module, and the abnormal work order identification module shown in fig. 9. The computer program constituted by the respective program modules causes the processor to execute the steps in the abnormal work order identification method of the respective embodiments of the present application described in the present specification.

For example, the computer device shown in fig. 10 may execute, by an image feature obtaining module in the abnormal work order recognition apparatus shown in fig. 9, obtaining a work order image in the currently reported work order, and obtaining a work order image feature corresponding to the work order image from the work order image. The computer equipment can execute resource transfer operation detection on the work order image according to the work order image characteristics through the first detection module to obtain a first detection result. The computer equipment can execute the blackening friend operation detection on the work order image according to the work order image characteristics through the second detection module to obtain a second detection result. The computer equipment can execute the resource transfer operation in the work order image of the current reported work order according to the first detection result and determine that the work order is the abnormal work order when the operation of the blacking friend exists in the work order image of the current reported work order according to the second detection result through the abnormal work order identification module.

In one embodiment, a computer device is provided, comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of the above-described abnormal work order identification method. Here, the steps of the abnormal work order identification method may be the steps in the abnormal work order identification method of each of the above embodiments.

In one embodiment, a computer readable storage medium is provided, storing a computer program that, when executed by a processor, causes the processor to perform the steps of the above-described abnormal work order identification method. Here, the steps of the abnormal work order identification method may be the steps in the abnormal work order identification method of each of the above embodiments.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a non-volatile computer-readable storage medium, and can include the processes of the embodiments of the methods described above when the program is executed. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present application. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims

1. An abnormal work order identification method, comprising:

2. The method of claim 1, further comprising, after said obtaining a work order image feature corresponding to said work order image from said work order image:

classifying the work order images according to the work order image characteristics to determine the category of the work order images;

the resource transfer operation detection of the work order image according to the work order image characteristics, and the obtaining of a first detection result comprises:

and when the type of the work order image is determined to be resource transfer, obtaining a first detection result of the resource transfer operation.

3. The method of claim 2, wherein the performing the resource transfer operation detection on the work order image according to the work order image feature and obtaining the first detection result comprises:

when the category of the work order image is determined to be chat, performing target detection on the work order image according to the work order image characteristics;

when the fact that the target information related to resource transfer is contained in the work order image is detected, a first detection result of resource transfer operation is obtained.

4. The method of claim 1, wherein prior to the detecting black friend operations on the work order image according to the work order image features, further comprising:

the performing blacking friend operation detection on the work order image according to the work order image characteristics, and obtaining a second detection result comprises:

and when the fact that the work order image contains target information associated with the black-pulling friend is detected, obtaining a second detection result of the operation of the black-pulling friend.

5. The method of any of claims 2 to 4, wherein the classifying the work order image according to the work order image features comprises:

inputting the work order image characteristics into a classification model determined through training, and classifying the work order images according to the classification model;

the classification model is obtained by obtaining a sample work order image carrying classification labels in a historical report work order and performing model training on a preset neural network model according to the sample work order image carrying the classification labels.

6. The method of claim 1, wherein the performing the resource transfer operation detection on the work order image according to the work order image feature and obtaining the first detection result comprises:

performing target detection on the work order image according to the work order image characteristics, and obtaining a first detection result of resource transfer operation when determining that the work order image contains target information related to resource transfer;

and performing target detection on the work order image according to the work order image characteristics, and obtaining a second detection result of the operation of the black-pulling friend when the work order image is detected to contain target information associated with the black-pulling friend.

7. The method according to claim 4 or 6, wherein after detecting that the work order image contains target information associated with a blacking friend, before obtaining a second detection result of the operation of the blacking friend, the method further comprises:

inputting the work order image into a character recognition model, judging that the work order image has the operation of the black-out friend when the work order image is determined to contain a preset keyword associated with the operation of the black-out friend according to the character recognition model, and obtaining a second detection result of the operation of the black-out friend.

8. The method of claim 3, 4 or 6, wherein the target detection of the work order image according to the work order image features comprises:

inputting the characteristics of the work order images into a target detection model determined through training, and carrying out target detection on the work order images according to the target detection model;

the target detection model is obtained by obtaining a sample work order image carrying target information labels in a historical report work order and performing model training on a preset neural network model according to the sample work order image carrying classification labels.

9. The method of any one of claims 1 to 4, wherein after acquiring the work order image in the currently reported work order, further comprising at least one of:

the first item is used for identifying the work order image and obtaining the image format of the work order image;

converting the image format of the work order image into a preset image format;

the second item is used for identifying the work order image and obtaining the image size of the work order image;

filtering the work order image with the image size not in accordance with the preset proportion, and adjusting the image size of the work order image to be the preset size;

thirdly, identifying the work order image to obtain an image channel of the work order image;

filtering the work order image of which the image channel comprises an alpha channel;

fourthly, identifying the work order image to obtain image pixel points of the work order image;

the obtaining of the work order image characteristics corresponding to the work order image according to the work order image includes: and obtaining the work order image characteristics corresponding to the work order image according to the image pixel points of the work order image.

10. An abnormal work order recognition apparatus, comprising:

11. A computer-readable storage medium, storing a computer program which, when executed by a processor, causes the processor to carry out the steps of the method according to any one of claims 1 to 9.

12. A computer device comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the method according to any one of claims 1 to 9.